In ACM hub cluster, 5050 and 6385 ports are exposed for Ironic and Ironic-Inspector.
These ports are used for managed baremetal nodes to connect the ACM hub cluster.
The IP address of the 5050/6385 ports will be the API VIP of the hub cluster.
In IPI installation, the API VIP is placed on one of the master nodes, so it works well.

However, in UPI installation, the API VIP is on an external loadbalancer, so the loadbalancer must have a loadbalancing configuration for 5050/6385 ports.
But currently there are no statements which describes this in our documentation.

I think we should have a section which mentions the required loadbalancing setting in the following document:

https://docs.redhat.com/en/documentation/red_hat_advanced_cluster_management_for_kubernetes/2.11/html/clusters/cluster_mce_overview#mce-network-configuration

The above document describes that 5050/6385 ports must be opened, but it's not enough.
If the ACM cluster is a UPI installation environment, we need to add a loadbalancing setting for the 5050/6385 ports in the external loadbalancer.
The following is an example configuration when the external loadbalancer is HAProxy.

listen ironic-api-6385
    bind *:6385
    mode tcp
    balance source
      server master-00 192.168.83.89:6385 check inter 1s
      server master-01 192.168.84.90:6385 check inter 1s
      server master-02 192.168.85.99:6385 check inter 1s

listen inspector-api-5050
    bind *:5050
    mode tcp
    balance source
      server master-00 192.168.83.89:5050 check inter 1s
      server master-01 192.168.84.90:5050 check inter 1s
      server master-02 192.168.85.99:5050 check inter 1s